I have a working shadow map system that uses 32-bit texture as the render target but I am trying to implement this using a depth-stencil texture to test performance / quality differences and am having some problems. I've been following the hardware shadow map example in the ATI SDK since i have a x1950Pro. I am using (D3DFORMAT)(MAKEFOURCC('D','F','1','6')) as the format and have tried (D3DFORMAT)(MAKEFOURCC('D','F','2','4')) with the same results. How are depth values written to the depth buffer? Is it just the projected Z value? In the sample here is how they export the position from the vertex shader: outPosition = mul( WorldTransformMatrix, inPosition ); outPosition.z *= RangeScale; where RangeScale is 1.0/FarClipZ and they use this depth for the comparison when rendering the scene: LightSpaceDepth = mul( WorldToLightTransformMatrix, Input.Position ).z * RangeScale; I've tried everything from following their sample, using the z values without the scaling and using the same z/w approach i use with my color target shadow maps. The shadow comparison always fails and the whole scene is in shadow except at very close distances to the light...